Update app.py

app.py

@@ -43,6 +43,21 @@ class GradioInference:
         self.tokenizer = AutoTokenizer.from_pretrained("csebuetnlp/mT5_multilingual_XLSum")
         self.model = AutoModelForSeq2SeqLM.from_pretrained("csebuetnlp/mT5_multilingual_XLSum")
 
+        self.llm_tokenizer = AutoTokenizer.from_pretrained("tiiuae/falcon-7b-instruct")
+        
+        self.pipeline = pipeline(
+            "text-generation", #task
+            model="tiiuae/falcon-7b-instruct",
+            tokenizer=self.llm_tokenizer,
+            torch_dtype=torch.bfloat16,
+            trust_remote_code=True,
+            device_map="auto",
+            max_length=1000,
+            do_sample=True,
+            top_k=10,
+            num_return_sequences=1,
+            eos_token_id=tokenizer.eos_token_id
+        )
 
     
     def __call__(self, link, lang, size, progress=gr.Progress()):
@@ -208,24 +223,7 @@ class GradioInference:
         ########################## PRUEBA LLM #################################
         from langchain import HuggingFacePipeline, PromptTemplate, LLMChain
         
-        llm_model = "tiiuae/falcon-7b-instruct"
-
-        llm_tokenizer = AutoTokenizer.from_pretrained(llm_model)
-        
-        pipeline = pipeline(
-            "text-generation", #task
-            model=llm_model,
-            tokenizer=llm_tokenizer,
-            torch_dtype=torch.bfloat16,
-            trust_remote_code=True,
-            device_map="auto",
-            max_length=1000,
-            do_sample=True,
-            top_k=10,
-            num_return_sequences=1,
-            eos_token_id=tokenizer.eos_token_id
-        )
-        llm = HuggingFacePipeline(pipeline = pipeline, model_kwargs = {'temperature':0})
+        llm = HuggingFacePipeline(pipeline = self.pipeline, model_kwargs = {'temperature':0})
         
         template = """
               Write a concise summary of the following text delimited by triple backquotes.